Fitting is not working on the triangular peak plot

[Marooz_Malik]

Dear Experts,
I tried to fit this histogram with double exponential + constant function. We also tried several other functions. But I am not able to do a combine fit of this histogram. I have also attached the code.
Can you please suggest me how can I fit this histogram?

fit.C (1.1 KB)

---

Please read tips for efficient and successful posting and posting code

ROOT Version: Not Provided
Platform: Not Provided
Compiler: Not Provided

---

[jonas]

Hi @Marooz_Malik,

what you have looks definitely like a Fermi function/sigmoid/cumulative Gaussian (all equivalent anyway), plus maybe a single exponential at the beginning (or even another sigmoid on top).

Here is how you could define a sigmoid plus exponential function:

// testFunc.C

double myfunc(double *x, double *p) {

    double u = p[0]; // uniform plateau hight
    double c = p[1]; // coefficient for exponential
    double tau = p[2]; // exponential parameter
    double mu = p[3]; // Gaussian position
    double sigma = p[4]; // Gaussian width
    
    using ROOT::Math::gaussian_cdf;
    return u * ((1-gaussian_cdf(x[0], sigma, mu)) + c * std::exp(-tau * x[0]));

}

void func() {

  auto *f1 = new TF1("f1", myfunc, 0, 30.0, 5);
  f1->SetTitle("Exponential plus cumulative Gaussian");
  f1->SetParameters(580000.0, 0.5, 1.0, 22.0, 3.0);
  f1->Draw();
  gPad->SaveAs("plot.png");
}

Hope this helps and gives you some ideas!
Jonas

[Marooz_Malik]

Dear Jonas,

I have tried this function and it is coming like this. I am not able to fit all the points.
What I want from this fitting is to extract some parameters such as slope of the constant region from 8 to 20.
I can somehow tweak the parameters so that fit looks good but the problem is that I have to do this for more than 500 histograms. Thus, I am looking for a general functions that will able to fit all the histograms which do not differ much. I am attaching some histos for reference.
Thanks

[Marooz_Malik]

can somebody help please ??

[Wile_E_Coyote]

It seems a simple “Exponential plus cumulative Gaussian” function is insufficient for describing your experimental data.
You need to talk to your colleagues and / or your supervisor about how to “model” them.

[Marooz_Malik]

thank you so much. will discuss and try to find a solution, then i can write it here.